Pulse Register Phonation in Diana Monkey Alarm Calls

Home , Diana monkey

View metadata, citation and similarPublished papers inat core.ac.ukThe Journal of the Acoustical Society of America, Vol. 113, Issue 5, 2003, p. 2919-2926 brought to you by CORE which should be used for any reference to this work provided by RERO DOC1 Digital Library

Pulse register phonation in Diana monkey alarm calls Tobias Riedea) Department of Psychology, 245 Uris Hall, Cornell University, Ithaca, New York 14853 Klaus Zuberbu¨hlerb) School of Psychology, University of St. Andrews, St. Andrews, Fife KY 16 9JU, Scotland, United Kingdom

The adult male Diana monkeys ͑Cercopithecus diana͒ produce predator-specific alarm calls in response to two of their predators, the crowned eagles and the leopards. The acoustic structure of these alarm calls is remarkable for a number of theoretical and empirical reasons. First, although pulsed phonation has been described in a variety of mammalian vocalizations, very little is known about the underlying production mechanism. Second, Diana monkey alarm calls are based almost exclusively on this vocal production mechanism to an extent that has never been documented in mammalian vocal behavior. Finally, the Diana monkeys’ pulsed phonation strongly resembles the pulse register in human speech, where fundamental frequency is mainly controlled by subglottal pressure. Here, we report the results of a detailed acoustic analysis to investigate the production mechanism of Diana monkey alarm calls. Within calls, we found a positive correlation between the fundamental frequency and the pulse amplitude, suggesting that both humans and monkeys control fundamental frequency by subglottal pressure. While in humans pulsed phonation is usually considered pathological or artificial, male Diana monkeys rely exclusively on pulsed phonation, suggesting a functional adaptation. Moreover, we were unable to document any nonlinear phenomena, despite the fact that they occur frequently in the vocal repertoire of humans and nonhumans, further suggesting that the very robust Diana monkey pulse production mechanism has evolved for a particular functional purpose. We discuss the implications of these findings for the structural evolution of Diana monkey alarm calls and suggest that the restricted variability in fundamental frequency and robustness of the source signal gave rise to the formant patterns observed in Diana monkey alarm calls, used to convey predator information.

I. INTRODUCTION mental frequency of the sound produced by the vocal folds is additionally directly related with the tension of the vocal fold The vocalizations of many mammals are the result of tissue ͑Titze, 1989, 1991͒. two distinct components: the oscillating vocal folds within Adult male Diana monkeys ͑Cercopithecus diana͒ pro- the larynx produce a primary acoustic signal, which then duce acoustically distinct alarm calls to two of their preda- undergoes a filtering process within the vocal tract where tors, the crowned eagle and the leopard ͑Zuberbu¨hler et al., various frequency bands are dampened to different degrees 1997; Zuberbu¨hler, 2000a͒. Playback experiments have ͑van den Berg, 1958; Fant, 1960; Titze, 1994; Owren and shown that nearby listeners respond to these alarm calls as if Linker, 1995͒. Basic vocal fold behavior can be described as the corresponding predator were present, suggesting that the following: Bernoulli forces cause the vocal folds ͑if close these calls inform nearby recipients about important ongoing to each other͒ to be sucked together, creating a closed air- events in the environment ͑Zuberbu¨hler et al., 1999; Zuber- space below the glottis. Continued subglottal air pressure bu¨hler, 2000b͒. Acoustically, the Diana monkeys’ alarm vo- from the lungs builds up underneath the closed folds. Once calizations consist of a bout of calls. Bouts vary in the num- this pressure becomes high enough, the folds are blown out- ber of calls from one to more than a dozen. Individual calls ward, thus opening the glottis and releasing a single ‘‘puff’’ are characterized by a highly stereotypic pulse pattern and ͑ ͒ of air ͑van den Berg, 1958͒. As the subglottal pressure in- calls are interspersed by short harmonic elements Fig. 1 . creases, two effects can be observed. First, the motion of the The single pulses within each call resemble a damped vocal folds becomes faster ͑demonstrated in computer mod- oscillation: a rapid, transient change in signal amplitude from els: Ishizaka and Flanagan, 1972; Steinecke and Herzel, a baseline value to a higher or lower value, followed by a 1995, and in vitro: Titze, 1989͒. Second, the sound pressure rapid return to the baseline value. Elsewhere, we showed that level increases ͑Gramming, 1988; Titze, 1994͒. The funda- the formant peak frequency and formant transition of the pulse elements is the single most important parameter to dif- ferentiate eagle versus leopard alarm calls ͑Riede and Zuber- a͒ Present address and address for correspondence: Tobias Riede, 315 Jordan bu¨hler, in press͒, suggesting that similar to human speech Hall, School of Medicine, Indiana University, Bloomington, IN 47405. Electronic mail: [email protected] sounds, some primate vocalizations convey important se- b͒Electronic mail: [email protected] mantic information by formant structures. Although research 2 completely adducted and a small vocal fold excursion ͑Hol- lien et al., 1977͒. The fundamental frequency is affected by different fac- tors in each of the three registers. In the modal register, the fundamental frequency is mainly determined by changes in vocal fold length and stiffness ͑Murry and Brown, 1971͒. Moreover, there is a positive correlation between vocal fold thickness ͑i.e., mass, length, and stiffness͒ and fundamental frequency ͓reviewed in Titze ͑1994͔͒. This relationship is absent in the pulse register ͑Hollien et al., 1969; Allen and Hollien, 1973͒. Instead, the fundamental frequency of the pulse register appears to be predominantly determined by changes in subglottal air pressure. To investigate the vocal production mechanism of the Diana monkey, we analyzed the relationship between call amplitude ͑a reliable estimator of subglottal pressure͒ and fundamental frequency. We predicted a positive relationship FIG. 1. Time domain and spectrogram of a leopard alarm bout, consisting of between these two parameters if Diana monkey alarm calls seven calls. Basic unit of the call is the pulse as shown in the ‘‘zoom in’’ picture of Fig. 2. are the product of the same source production mechanism that is responsible for the human pulse register. on the structural evolution of animal vocalizations is not new A second aim of this study was to investigate the role of ͑Morton, 1977͒, comparatively little is still known about how nonlinear phenomena in the vocalizations of Diana monkey natural and sexual selection affected the acoustic structure of alarm calls. Nonlinear phenomena are relevant in this context primate alarm calls ͑Zuberbu¨hler, 2003͒. Here, we provide a because they can be directly related to events at the laryngeal detailed acoustic analysis of the source characteristics of Di- source. Several lines of research suggest that nonlinear phe- ana monkey alarm calls to elucidate the adaptive significance nomena are common and ubiquitous in mammalian vocalization behavior ͑Wilden et al., 1998; Mergell et al., 1999; and physiological constraints of this remarkable vocalization. ͒ Human speech sounds can be produced using three dif- Riede et al., 1997, 2000; Fischer et al., 2000 . Phenomena ferent registers. A register can be described by the frequency such as frequency jumps, subharmonics, biphonation, and range covered and by the specific mode of vocal fold behav- deterministic chaos are commonly observed, usually the re- ior by which it is produced ͑e.g., Hollien, 1974; Titze, 1994; sult of deviations from the regular harmonic vibration pattern Svec et al., 1999͒. Although each register covers a certain of the vocal folds, such as nonsynchronously oscillating left and right vocal folds or simultaneously oscillating horizontal frequency range, neighboring registers overlap significantly. ͑ Normal speech is delivered in the so-called modal ͑or chest͒ and vertical components of the vocal folds e.g., Herzel ͑ ͒ et al., 1994; Berry, 2001; Berry et al., 1994; Steinecke and register fundamental frequency range 100–300 Hz . Hu- ͒ mans are also capable to produce speech using either the Herzel, 1995; Tigges et al., 1997; Neubauer et al., 2001 . falsetto ͑or flagolet͒ register ͑fundamental frequency Ͼ300 The two combined approaches are likely to yield important Hz͒ or the pulse ͑or vocal fry͒ register ͓fundamental fre- insights into the sound production mechanism underlying quency Ͻ100 Hz ͑Blomgren et al., 1998͔͒. Recent studies male Diana monkey alarm calls. suggest the existence of a separate fourth register, i.e., the vocal-ventricular phonation mode, like pulse register cover- II. MATERIAL AND METHODS ing the frequency range below 100 Hz but unlike pulse reg- A. Study site and subjects ister involving the ventricular folds ͑‘‘false folds’’͒ into the mode of production ͑Fuks et al., 1998; Lindestad et al., Data were collected in an approximately 40-km2 study- 2001͒. According to this terminology we used the term area of primary rain forest surrounding the Centre en ‘‘pulse register’’ to describe the Diana monkey calls, because Recherche d’Ecologie ͑University of Cocody, Abidjan͒ re- these vocalizations strongly resemble the pulse register of search station ͑5°50ЈN, 7°21ЈW͒ in the Taı¨ National Park, humans ͓see Blomgren et al. ͑1998͒ for a review͔. Coˆte d’Ivoire, between June 1994 and June 1997. Seven Pulse register differs in acoustical, physiological, and monkey species are regularly observed in the area: the west- perceptual characteristics from other phonation types ͓re- ern red colobus ͑Colobus badius͒, the western black-and- viewed in Gerratt and Kreiman ͑2001͔͒. Vocal fold vibration white colobus ͑Colobus polykomos͒, the olive colobus ͑Pro- during pulse register is characterized by glottal pulses of al- colobus verus͒; the Diana monkey ͑Cercopithecus diana͒, ternating amplitudes or by irregular trains of pulses ͑Hollien the lesser white-nosed monkey ͑Cercopithecus petaurista͒, and Michel, 1968͒. The vocal fold length is shorter for the the Campbell’s monkey ͑Cercopithecus campbelli͒, and the pulse register than for even the lowest frequency of phona- sooty mangabey ͑Cercocebus torquatus͒. Diana monkey tion in the modal register ͑Hollien et al., 1969͒. The vocal groups typically consist of about 20–25 individuals with one fold vibratory pattern of the pulse register in humans exhibits adult male and several adult females with their offspring. a very short open period ͑probably less than 25% of the Groups occupy stable home ranges of approximately 60 ha. entire cycle͒ and a very long period where the vocal folds are Diana monkeys eat primarily fruit and insects and they are 3 found at all levels of the forest but prefer the main upper canopy. None of the animals were habituated to human presence. However, all data were so that the animals were un- aware of the observer’s presence.

B. Recordings and acoustic analysis We recorded Diana monkey alarm vocalizations given in response to playbacks of African leopard ͑Panthera pardus͒ and crowned eagle ͑Stephanoaetus coronatus͒ vocalizations, using a Sony WMD6C tape recorder and a Sennheiser mi- crophone ͑ME88 head with K3U power module͒ on 90-min type IV metal tapes. The frequency response of the micro- phone ͑40 Hz to 20 kHz; Ϯ2.5 dB͒ and the tape recorder ͑40 Hz to 14 kHz, Ϯ3 dB; distortion of 0.1%; signal-to-noise- ratio of 57 dB͒ are flat and within the frequency range of analysis. Playbacks of predator vocalizations were conducted randomly throughout the day, usually between 08:00 and 17:00 GMT. Daytime therefore cannot explain the differences in the vocal patterns. All recordings were made at distances of about 50 m from the focal animal, i.e., the adult male of a Diana monkey group. Individual distances varied randomly across trials and therefore cannot explain the patterns. Male alarm calls transmit to about 700 m, i.e., sound attenuation at short distances was unlikely to have affected the acoustic variables. The study area contained between 40 and 80 different groups of Diana monkeys with one adult male each. Because we did not know the exact location of these groups’ home ranges, we selected ten different groups for experimental playbacks that were located at least 1 km apart from each other, which guaranteed that data came from ten different groups, i.e., were independent. This data set resulted in a set of 25 eagle alarm bouts ͑5 bouts from 5 different males͒ and 25 leopard alarm bouts ͑5 bouts from 5 different males͒. We digitized all recordings at a 16-bit quan- tization and a 44-kHz sampling rate using Signalize soft- FIG. 2. Time series of pulses of five different individuals. The pulse is the ware. We performed signal analysis on a PC using the signal basic acoustic unit in the alarm call; it is defined as a rapid, transient change processing software HYPERSIGNAL-Macro™ using a in the amplitude of the signal from a baseline value to a higher or lower DSP32C PC System Board. We completed the spectro- value, followed by a rapid return to the baseline value. Arrows point to the graphic analysis by using 512-point fast Fourier transforms, first or second peak in the pulse waveform. with 75% frame overlapping, a 44-kHz sampling frequency, and a Hanning window. To avoid aliasing effects we low- passed filtered all calls at 22 kHz. fined as cycle-to-cycle variability in the fundamental frequency ͑Titze, 1994͒. Fundamental frequency ranges between 8.3 and 24 ms ͑meanϮSD 16.1Ϯ2.0͒ in eagle alarm C. Call parameters calls and between 13.3 and 29.9 ms ͑meanϮSD 17.4Ϯ2.4͒ in A male Diana monkey alarm vocalization consists of leopard alarm calls ͑Riede and Zuberbu¨hler, in press͒. Call one to many calls per bout ͑Fig. 1͒. The basic acoustic unit duration and jitter ranges in eagles and leopard alarm calls within a call is the pulse, defined as a rapid, transient change between 6.4% and 9.2% ͑Riede and Zuberbu¨hler, in press͒. in the amplitude of the signal from a baseline value to a In this study we investigate the development of the two pa- higher or lower value, followed by a rapid return to the base- rameters maximum amplitude of a pulse and fundamental line value, resembling a damped oscillation ͑Fig. 2͒. frequency within a call. Both parameters were normalized Pulse duration is measured as the interval between the within calls. For the correlation between maximum ampli- onset of a pulse to the onset of the subsequent pulse. Funda- tude of a pulse and fundamental frequency we considered the mental frequency is defined as the inverse ͑i.e., 1/pulse du- means of the pulses at position 0%, 25%, 50%, 75%, and ration in Hz; pulse duration measured in seconds͒. Through- 100% within the call. Only calls with more than ten pulses out the paper we use the term ‘‘fundamental frequency’’ to and with very low background noise level, i.e., a high signal- refer to the inverse value of ‘‘pulse duration’’ in the wave- to-noise ratio, were considered, resulting in a data set of 10 form. We quantified the variation of the fundamental fre- leopard and 21 eagle alarm calls, respectively. quency within calls by the parameter within-call jitter, de- To test if pulse time series are the result of individual- 4

FIG. 4. Time domain of a call. Note the increase in amplitude toward the middle of the call and the amplitude decreases toward the end.

Even maximum amplitude of pulses depended on the position of the pulse within a call. Pulses in the middle of the call were louder than those at the beginning or the end of the call ͑Fig. 4͒. Figure 5 summarizes the development of maximum amplitude of pulses within calls of leopard alarm calls (Nϭ10 calls͒ and eagle alarm calls (Nϭ21 calls͒. Comparing both parameters, there is a suggestive positive correlation between fundamental frequency and maximum amplitude of pulses within a call in leopard alarm calls ͑Pearson, Nϭ5, rϭ0.8, Pϭ0.1) and there is a signiﬁcant positive correlation between fundamental frequency and maximum amplitude of pulses in eagle alarm calls ͑Pearson, Nϭ5, rϭ0.96, Pϭ0.0089). For the correlation the ﬁve mean values, as shown in Figs. 3 and 4, have been used.

FIG. 3. Fundamental frequency over syllables in ͑a͒ leopard alarm calls (Nϭ10 syllables͒ and ͑b͒ in eagle alarm calls (Nϭ21 syllables͒. Each data point in the diagram represents the mean Ϯ standard deviation of the rela- tive fundamental frequency within Nϭ10 calls ͑leopard alarm͒ and Nϭ21 calls ͑eagle alarm͒. Since calls are of different duration, i.e., they consist of a different number of pulses, call duration was standardized. The five pulses on positions 0%, 25%, 50%, 75%, and 100% of the total number of pulses within a call were considered for the graphs. specific patterns, cross correlations between pulse time series were undertaken. Five pulses were selected from each of five calls, cut and saved as a text compatible ASCII file. In NCSS 2001 statistical software single cross correlations were run ͑a͒ on the within-call level, ͑b͒ the between-call and within- individual level, and ͑c͒ on the between-individual level. Fi- nally, we were interested in the occurrence of nonlinear phenomena ͑frequency jumps, subharmonics, biphonation, deterministic chaos͒ in the alarm calls. For this purpose we inspected the call spectrograms visually for consistency of the pulse pattern, using a data set of 50 calls plus an additional data set of 100 calls from other individuals.

III. RESULTS A. Fundamental frequency versus maximum amplitude of a pulse FIG. 5. Maximum amplitude of pulses ͑a͒ in leopard alarm calls (Nϭ10 Fundamental frequency depended on the position of the calls͒ and ͑b͒ in eagle alarm calls (Nϭ21 calls͒. Each data point within the pulse within the call. Fundamental frequency was lower at diagram represents the mean Ϯ standard deviation of the maximum ampli- the beginning and at the end of the call than in the middle of tude of pulses of Nϭ10 calls ͑leopard alarm͒ and Nϭ21 calls ͑eagle alarm͒. the call. Figure 3 summarizes the development of fundamen- Since calls are of different duration, i.e., they consist of a different number ϭ ͒ of pulses, call duration was standardized. The five pulses on positions 0%, tal frequency over a call of leopard alarm calls (N 10 calls 25%, 50%, 75%, and 100% of the total number of pulses within a call were and eagle alarm calls (Nϭ21 calls͒. considered for the graphs. 5 is a short harmonic element, a sound probably uttered during inspiration. The pulse pattern is very robust, being not interrupted by other phonation types ͑vibration modes of the source͒. In 150 different calls produced by more than a dozen different males we did not find any other than the pulse pattern. The fundamental frequency of male Diana monkey alarm calls ranges between 33 and 120 Hz ͑Riede and Zuberbu¨hler, in press͒ similar to fundamental frequency ranges of pulsed phonation in Felidae ͓F0 in purring between 10 and 45 Hz ͑Peters and Tonkin-Leyhausen, 1999͔͒ and humans ͓F0 in pulse register between 10 and 90 Hz ͑Henton and Bladon, 1988͔͒. Although similar patterns were found for instance in felids ͑Peters and Tonkin-Leyhausen, 1999͒ or humans ͑Titze, 1994͒ unlike to male Diana monkey, cats do produce all kinds of other vocalizations ͑Peters, FIG. 6. Cross correlations of pulse time series on three levels. WS—within 1981͒ and in humans the occasional occurrence of a subhar- calls, BS&WI—between calls and within individuals, BI—between indi- monic regime within a pulse register utterance is reported viduals. Cross correlation values can be considered as similarity indexes ͑ ͒ between two time series, saying the higher the cross correlation coefficient Titze, 1994 . the higher the similarity. Within calls the similarity between pulses is high- The aim of the present study was to investigate the est. acoustic characteristics of pulsed phonation in Diana monkeys. The sound production mechanism in this species is of B. Similarity in the pulse waveform particular interest since it has been shown that formant characteristics of a single pulse conveys important information to The waveform of a pulse varied within and between nearby listeners about ongoing predation events ͑Zuber- individuals. For instance, the maximum amplitude of a pulse bu¨hler, 2000b; Riede and Zuberbu¨hler, in press͒. Our data can be consistent within the first single cycle of a pulse or confirmed the very narrow range in fundamental frequency alternatively, within one of the later cycles ͑indicated by ar- in the Diana monkey pulse register, suggesting very limited rows in Fig. 2͒. Cross correlations between single pulses vocal fold adjustments. In a given adjustment of the vocal showed an individual specific pattern, delivering highest folds ͑i.e., a given length and tension͒, which is not changed cross correlation values within calls, being less similar be- during a single utterance, fundamental frequency seems to be tween calls within individuals, and being least between indi- exclusively regulated by the one variable—subglottal pres- viduals ͑Fig. 6͒. The differences between conditions were sure ͑Murry and Brown, 1971͒. This stands in contrast to the significant (N ϭ25, N ϭ25, N ϭ25, Fϭ89.2, PϽ0.001) 1 2 3 modal phonation type, where the fundamental frequency is with posthoc comparisons showing that all means differed controlled mainly by vocal fold tension. Subglottal pressure from one another. As illustrated in Fig. 6, the means of the ͑ has been found to correlate both with fundamental frequency two ‘‘within individual’’ conditions within calls and be- ͑ tween calls͒ were closer to each other than the ‘‘between Ishizaka and Flanagan, 1972; Steinecke and Herzel, 1995; ͒ ͑ individual’’ condition to each of the other two conditions Titze, 1989 and with signal amplitude Gramming, 1988; ͒ ͑Fig. 6͒. Titze, 1994 . In male Diana monkeys, we found a positive correlation between the fundamental frequency and the maximum am- C. Nonlinear phenomena plitude of a pulse. Since signal amplitude is mainly con- We investigated the calls for the occurrence of nonlinear trolled by subglottal pressure ͑Gramming, 1988; Titze, phenomena. Pulses occurred in a very regular pattern, i.e., 1994͒, we conclude that fundamental frequency in male Di- visual inspection of the spectrograms delivered no deviations ana monkey alarm calls is similarly controlled by subglottal from the pulse pattern, suggesting a rigid and not deviating pressure. Male Diana monkeys differ from humans in that vibration pattern of the oscillating system ͑of the sound they apparently do not switch to a higher register to produce source͒. This was true for the whole data set of 50 calls. vocalizations with higher fundamental frequencies. Male Di- Even in the additional data set of 100 additional calls no ana monkeys, it appears, are constrained by a pulsed phona- nonlinear phenomena were discovered. tion mechanism whose fundamental frequencies are entirely regulated by subglottal pressure. IV. DISCUSSION A. The evolution of Diana monkey alarm calls The acoustic structure of male Diana monkey alarm Previous work has shown that the fundamental fre- calls is remarkable. These vocalizations consist of trains of quency of mammalian vocalizations tends to covary with loud and low-pitched calls that carry over long distances of context relevant aspects, like individual identity, sex or de- up to a kilometer through dense tropical forest habitat. A gree of arousal ͓reviewed in Tembrock ͑1996͔͒. If, however, pulse of 8- to 30-ms duration is the basic unit of male Diana the primary signal is rigid, repetitive and broadband and monkey alarm vocalizations. Up to 30 pulses are associated shows little variability in its most important parameter fun- to a call. Several calls build a bout. Between two calls there damental frequency, as it is the case in the Diana monkey, 6 TABLE I. Examples of pulsed utterances in other species than Diana monkeys, giving the name of the utterance, the extent of occurrence, and the reference.

Species Call type Description Reference

Primates Human, Homo Creaky voice Voluntarily, end-of Reviewed in sapiens utterance Henton and Bladon phenomenon or as ͑1988͒ a pathological voice

Chacma baboon, Wahoo End of call with Fischer et al. ͑2002͒ Papio ursinus some few pulses Gelada baboon, Richman ͑1976͒ Theropithecus gelada

Pigtailed macaque, Intention notes, A group of calls Grimm ͑1967͒ Macaca nemestrina inspiratory note, subsummarized as vibrato growl, bark ‘‘harsh sounds’’

Squirrel monkey, Girren, churr A separate call within Winter ͑1969͒͑Fig. 9͒; Saimirisciureus the repertoire Ploog et al. ͑1975͒ Red howling Roars In the climax of the Scho¨n Ybarra ͑1986͒ monkey, Alouatta roar ͑p. 209͒ seniculus

Other mammals Several felidae Purring Peters ͑1981͒ Koala, Bellow Seemingly the Smith ͑1980͒͑p. 21, Phascolarctos whole utterance is Figs. 9 and 10͒ cinereus pulsed

Cetaceans Clicks Part of the sonar Au ͑1993͒ system or of other repertoire then this might provide a reliable and fruitful basis for the of the source, i.e., to engage in articulatory maneuvers. Re- evolution of more sophisticated vocal tract performance. In- search focusing on the signal production mechanisms will be deed, other work has shown that formant modulation plays necessary to determine the general evolutionary trends that the most important role in the acoustic differentiation of were likely to have affected Diana monkey vocal behavior. eagle and leopard alarm calls in male Diana monkeys In recent years it became more evident that the mamma- ͑Zuberbu¨hler, 2000b; Riede and Zuberbu¨hler, in press͒. Di- lian larynx can be considered as a nonlinear system; several ana monkeys manage to filter the primary source signal pro- studies showed that sudden changes in the vibration mode of duced by the vocal folds in their vocal tracts to product the vocal folds are more the normal picture rather than an acoustically distinct eagle and leopard alarm calls. Because exception ͑Wilden et al., 1998; Riede et al., 2000͒. Those of their broad bandwidth, pulses are particularly well suited studies are contrasted by the findings presented here, show- to picture the resonance characteristics of the vocal tract and ing that male Diana monkey alarm call pulsed phonation was serve as acoustic raw material for filtering effects in the vo- free of any interruption of the pulse pattern. Anatomical data cal tract. on the male Diana monkey larynx will be necessary to illu- Interestingly, human singing tutelage is often based on minate the evolution of this special kind of vocalizations. using pulse register phonation as an exercise to ‘‘tune’’ the Possibly, anatomical adaptations make the vocal folds ‘‘spe- vocal tract ͑Miller et al., 1997͒. By singing a particular cial’’ for this kind of vibration behavior, leading to the ob- vowel in pulse register the trainee can examine his or her served highly stabilized vibration pattern. vocal tract performance. Once successful, the trainee switches back to the actual singing voice while maintaining B. Pulse phonation—an exclusive vocal pattern in male Diana monkeys? the vocal tract configuration. This exercise should enable the performer to tune the formants and fundamental frequency Table I reviews studies that mention pulsed phonation as optimally. part of a species repertoire or present spectrograms, which This example illustrates that the simple and robust pulse suggest the occurrence of pulsed vocal utterances. signal is physiologically easy to produce, but is insufficient In land mammals, the felids seem to be the most inten- as a source of acoustic variation to be useful to convey con- sively studied group that show pulsed phonation ͑Peters and textual information. Instead, phonation based on a pulse sig- Tonkin-Leyhausen, 1999͒. The time series of purring, how- nal is likely to favor the evolution of vocal tract characteris- ever, appears different from Diana monkey alarm calls and tics that enable the caller to engage in sophisticated molding humans pulse register ͑Fig. 7͒. In humans, pulse register 7 ͑e.g., Abel, 1972; Buus and Florentine, 1985͒. Assuming the gap detection ability of nonhuman primates is comparable to that of humans, the pulse duration of male Diana monkey alarm calls lies comfortably above that threshold, suggesting the pulsed structure of Diana monkey alarm calls is per- ceived as an attention attracting structure. Psychophysical experiments manipulating the number of pulses might be suitable to determine the minimum amount of information necessary for a Diana monkey to discern eagle from leopard alarm calls or to identify calls as those of a conspecific.

V. CONCLUSION To investigate the mechanisms of sound production in Diana monkeys, eagle and leopard alarm calls from ten different Diana monkey males were digitized and subjected to spectrographic analysis. Results showed that the fundamental frequency of these calls ranged between 33 and 120 Hz, comparable to the human pulse register, which tends to range between 10 and 90 Hz. Jitter was very small and did not vary significantly between individuals or alarm call type. Nonlin- ear phenomena were virtually absent in male Diana monkey vocalization and pulses were not interrupted by any other vibration modes of the vocal folds. Over the entire calls, fundamental frequency was low at the beginning and at the end of the syllable and highest in the middle of the call, while the amplitudes of pulses increase towards the middle and then decrease toward the end of the call, indicating that fundamental frequency and maximum amplitude of a pulse were correlated, which suggested that the fundamental frequency in Diana alarm calls is controlled by subglottal pressure rather than vocal fold stiffness changes. The pulsed phonation in male Diana monkey alarm call, therefore, appeared FIG. 7. Time series of pulsed phonation in different land mammals. ͑d͒ Domestic cat, ͑c͒ leopard, ͑b͒ human creaky voice, and ͑a͒ Diana monkey. to be a special adaptation delivering a robust source broadband signal for subsequent vocal tract filtering. does not seem to play an important role in everyday speech. Instead it is considered an ‘‘end-of-utterance’’ phenomenon, ACKNOWLEDGMENTS indicating that pulses often occur at the end of words or We thank Michael Owren and two anonymous reviewers ͑ sentences, and in men more often than in women Henton for important comments on an earlier version of the manu- ͒ and Bladon, 1988 . However, no studies have dealt in any script. This work was supported by a fellowship within the depth with the incidence of vocal fry in natural, unprovoked Postdoc Program of the German Academic Exchange Service ͑ ͒ conditions in humans Henton and Bladon, 1988 . For other ͑DAAD͒ to TR. species this kind of a ‘‘single mode’’ source signal was apparently not yet described. In contrast to all these findings, Abel, S. M. ͑1972͒. ‘‘Discrimination of temporal gaps,’’ J. Acoust. Soc. Am. our results suggest that in male Diana monkeys pulse register 52, 519–524. Allen, E., and Hollien, H. ͑1973͒. ‘‘A laminagraphic study of pulse ͑vocal is the main mode at the source. fry͒ phonation,’’ Folia Phoniatr. 25, 241–250. Au, W. W. L. ͑1993͒. The Sonar of the Dolphin ͑Springer Verlag, New York͒. C. Implications for the receiver Berry, D. A. ͑2001͒. ‘‘Mechanisms of modal and nonmodal phonation,’’ J. Phonetics 29, 431–450. Animal alarm calls have often been shown to be repeti- Berry, D. A., Herzel, H., Titze, I. R., and Krischer, K. ͑1994͒. ‘‘Interpreta- tive, i.e., consisting of short and similarly structured seg- tions of biomechanical simulations of normal and chaotic vocal fold os- ments which are repeated, for instance in the vervet monkey cillations with empirical eigenfunctions,’’ J. Acoust. Soc. Am. 95,3595– snake alarm calls ͑Owren, 1990͒. Studies on the psychologi- 3604. Blomgren, M., Chen, Y., Ng, M. L., and Gilbert, H. R. ͑1998͒. ‘‘Acoustic, cal value of visual and acoustic signals have shown that re- aerodynamic, physiologic, and perceptual properties of modal and vocal peating short elements to form a signal are the most effective fry registers,’’ J. Acoust. Soc. Am. 103, 2649–2658. way to attract attention and alert others ͓reviewed in Brad- Bradbury, J. W., and Vehrencamp, S. L. ͑1998͒. Principles of Animal Com- ͑ ͒ bury and Vehrencamp ͑1998͔͒. Diana monkey alarm call munication Sinauer, Sunderland . Buus, S., and Florentine, M. ͑1985͒. ‘‘Gap detection in normal and impaired pulses are short and repetitive. It is interesting to note that listeners: The effect of level and frequency,’’ in Time Resolution in Audi- silent gaps of about 5 ms can still be detected by humans tory Systems, edited by A. Michelsen ͑Springer, New York͒. 8 Fant, G. ͑1960͒. Acoustic Theory of Speech Production ͑Mouton, The cation signals of mammals: friendly close range vocalizations in felidae Hague͒. ͑Carnivora͒,’’ J. Mammalian Evol. 6, 129–159. Fischer, J., Hammerschmidt, K., Cheney, D. L., and Seyfarth, R. M. ͑2002͒. Ploog, D., Hupfer, K., Ju¨rgens, U., and Newman, J. D. ͑1975͒. ‘‘Neuroetho- ‘‘Acoustic features of male baboon loud calls: influences of context, age, logical studies of vocalization in squirrel monkeys with special reference and individuality,’’ J. Acoust. Soc. Am. 111, 1465–1474. to genetic differences of calling in two subspecies,’’ in Growth and De- Fuks, L., Hammarberg, B., and Sundberg, J. ͑1998͒. ‘‘A self-sustained velopment, edited by M. A. B. Brazier ͑Raven, New York͒. vocal-ventricular phonation mode: acoustical, aerodynamic and glot- Richman, B. ͑1976͒. ‘‘Some vocal distinctive features used by gelada mon- tographic evidences,’’ TMH-QPRS 3, 49–59. keys,’’ J. Acoust. Soc. Am. 60, 718–724. Gerratt, B., and Kreiman, J. ͑2001͒. ‘‘Toward a taxonomy of nonmodal Riede, T., and Zuberbu¨hler, K. ͑in press͒. ‘‘The relationship between acous- phonation,’’ J. Phonetics 29, 365–381. tic structure and semantic information in Diana monkey alarm calls.’’ J. Gramming, P. ͑1988͒. ‘‘The phonetogram: An experimental and clinical Acoust. Soc. Am. study,’’ Malmoe Sweden, Dept. Otolaryngology, Univ. of Lund. Riede, T., Wilden, I., and Tembrock, G. ͑1997͒. ‘‘Subharmonics, biphona- Grimm, R. J. ͑1967͒. ‘‘Catalogue of sounds of the pigtailed macaque tions, and frequency jumps—common components of mammalian vocal- ͑Macaca nemestrina͒,’’ J. Zool. Lond. 152, 361–373. ization or indicators for disorders,’’ Z. Saügetierkunde 62, 198–203. Henton, C., and Bladon, A. ͑1988͒. ‘‘Creak as a sociophonetic marker,’’ in Riede, T., Herzel, H., Mehwald, D., Seidner, W., Trumler, E., Bo¨hme, G., Language, Speech, and Mind: Studies in Honour of Victoria A. Fromkin, and Tembrock, G. ͑2000͒. ‘‘Nonlinear phenomena in the natural howling edited by L. M. Hyman and C. N. Li ͑Routledge, London͒, pp. 3–29. of a dog-wolf mix,’’ J. Acoust. Soc. Am. 108, 1435–1442. Herzel, H., Berry, D., Titze, I. R., and Saleh, S. ͑1994͒. ‘‘Analysis of vocal Schoen Ybarra, M. A. ͑1986͒. ‘‘Loud calls of adult male red howling mon- disorders with methods from nonlinear dynamics,’’ J. Speech Hear. Res. keys ͑Alouatta seniculus͒,’’ Folia Primatol ͑Basel͒ 47, 204–216. 37, 1008–1019. Smith, M. ͑1980͒. ‘‘Behavior of the Koala, Phascolarctos cinereus ͑Gold- Hollien, H. ͑1974͒. ‘‘On vocal registers,’’ J. Phonetics 2, 125–143. fuss͒, in captivity, III. Vocalizations,’’ Aust. Wildl. Res. 7, 13–34. Hollien, H., and Michel, J. ͑1968͒. ‘‘Vocal fry as a phonational register,’’ J. Steinecke, I., and Herzel, H. ͑1995͒. ‘‘Bifurcations in an asymmetric vocal Speech Hear. Res. 11, 600–604. fold model,’’ J. Acoust. Soc. Am. 97, 1874–1884. Hollien, H., Damste´, H., and Murry, T. ͑1969͒. ‘‘Vocal fold length during Svec, J. G., Schutte, H. K., and Miller, D. G. ͑1999͒. ‘‘On pitch jumps vocal fry phonation,’’ Folia Phoniatr. 21, 257–265. between chest and falsetto registers in voice: Data from living and excised Hollien, H., Girard, G. T., and Coleman, R. F. ͑1977͒. ‘‘Vocal fold vibratory human larynges,’’ J. Acoust. Soc. Am. 106, 1523–1531. patterns of pulse register phonation,’’ Folia Phoniatr. 29, 200–205. ͑1972͒. Tembrock, G. ͑1996͒. Akustische Kommunikation der Saügetiere. Wissen- Ishizaka, K., and Flanagan, J. L. ͑1972͒. ‘‘Synthesis of voiced sounds from schaftl ͑Buchgesell, Darmstadt͒. a two-mass model of the vocal cords,’’ Bell Syst. Tech. J. 51, 1233–1268. Tigges, M., Mergell, P., Herzel, H., Wittenberg, T., and Eysholdt, U. ͑1997͒. Lindestad, P., Soedersten, M., Merker, B., and Granqvist, S. ͑2001͒. ‘‘Voice ‘‘Observation and modeling of glottal biphonation,’’ Acustica 83, 707– source characteristics in Mongolian ‘throat singing’ studied with high- 714. speed imaging technique, acoustic spectra, and inverse filtering,’’ J. Voice Titze, I. R. ͑1989͒. ‘‘On relation between subglottal pressure and fundamen- 15, 78–85. tal frequency in phonation,’’ J. Acoust. Soc. Am. 85, 901–906. Mergell, P., Fitch, W. T., and Herzel, H. ͑1999͒. ‘‘Modeling the role of Titze, I. R. ͑1991͒. ‘‘Mechanisms underlying the control of fundamental non-human vocal membranes in phonation,’’ J. Acoust. Soc. Am. 105, frequency,’’ in Vocal Fold Physiology: Acoustic, Perceptual and Physi- 2020–2028. ological Aspects of Voice Mechanisms, edited by J. Gauffin and B. Ham- Miller, D. G., Sulter, A. M., Schutte, H. K., and Wolf, R. F. ͑1997͒. ‘‘Com- marberg ͑Singular, San Diego͒, pp. 129–138. parison of vocal tract formants in singing and nonperiodic phonation,’’ J. Titze, I. R. ͑1994͒. Principles of Voice Production ͑Prentice Hall, Engle- Voice 11, 1–11. wood Cliffs, NJ͒. Morton, E. S. ͑1977͒. ‘‘On the occurrence and significance of motivation— van den Berg, J. ͑1958͒. ‘‘Myoelastic-aerodynamic theory of voice produc- structural rules in some bird and mammal sounds,’’ Am. Nat. 111,855– tion,’’ J. Speech Hear. Res. 1, 227–244. 869. Wilden, I., Herzel, H., Peters, G., and Tembrock, G. ͑1998͒. ‘‘Subharmonics, Murry, T., and Brown, W. S., Jr. ͑1971͒. ‘‘Subglottal air pressure during two biphonation, and deterministic chaos in mammal vocalization,’’ Bioacous- types of vocal activity: vocal fry and modal phonation,’’ Folia Phoniatr. tics 9, 171–196. 23, 440–449. Winter, P. ͑1969͒. ‘‘Dialects in squirrel monkeys: vocalization of the roman Neubauer, J., Mergell, P., Eysholdt, U., and Herzel, H. ͑2001͒. ‘‘Spatio- arch type,’’ Folia Primatol ͑Basel͒ 10, 216–229. temporal analysis of irregular vocal fold oscillations: biphonation due to Zuberbu¨hler, K. ͑2000a͒. ‘‘Causal knowledge of predators’ behavior in wild desynchronization of spatial modes,’’ J. Acoust. Soc. Am. 110, 3179– Diana monkeys,’’ Anim. Behav. 59, 209–220. 3192. Zuberbu¨hler, K. ͑2000b͒. ‘‘Referential labeling in Diana monkeys,’’ Anim. Owren, M. J. ͑1990͒. ‘‘Acoustic classification of alarm calls by veret mon- Behav. 59, 917–927. keys ͑Cercopithecus aethiops),’’ J. Acoust. Soc. Am. 101, 2951–2963. Zuberbu¨hler, K. ͑2003͒. ‘‘The effects of natural and sexual selection on the Owren, M., and Linker, C. D. ͑1995͒. ‘‘Some analysis techniques that may evolution of guenon loud calls,’’ in The Guenons, edited by M. Glenn and be useful to acoustic primatologists,’’ in Current Topics in Primate Vocal M. Cords ͑Plenum, New York͒. Communication, edited by E. Zimmermann, J. Newman, and U. Ju¨rgens Zuberbu¨hler, K., Cheney, D. L., and Seyfarth, R. M. ͑1999͒. ‘‘Conceptual ͑Plenum, New York͒, pp. 1–27. semantics in a non-human primate,’’ J. Comp. Psychol. 113,33–42. Peters, G. ͑1981͒. ‘‘Das Schnurren der Katzen,’’ Saügetierkundl. Mitteilun- Zuberbu¨hler, K., Noe, R., and Seyfarth, R. M. ͑1997͒. ‘‘Diana monkey long- gen 29, 30–37. distance calls: messages for conspecifics and predators,’’Anim. Behav. 53, Peters, G., and Tonkin-Leyhausen, B. A. ͑1999͒. ‘‘Evolution of communi- 589–604.